@speech-sdk/core 0.21.0 → 0.23.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +69 -17
- package/dist/conversation/stitch.d.ts +2 -0
- package/dist/conversation/stitch.d.ts.map +1 -1
- package/dist/conversation/stitch.js +14 -34
- package/dist/conversation/stitch.js.map +1 -1
- package/dist/conversation/types.d.ts +2 -0
- package/dist/conversation/types.d.ts.map +1 -1
- package/dist/derive-timestamps.d.ts.map +1 -1
- package/dist/derive-timestamps.js +2 -2
- package/dist/derive-timestamps.js.map +1 -1
- package/dist/errors.d.ts +40 -0
- package/dist/errors.d.ts.map +1 -1
- package/dist/errors.js +125 -1
- package/dist/errors.js.map +1 -1
- package/dist/generate-conversation.d.ts +5 -2
- package/dist/generate-conversation.d.ts.map +1 -1
- package/dist/generate-conversation.js +87 -7
- package/dist/generate-conversation.js.map +1 -1
- package/dist/generate-speech.d.ts +4 -1
- package/dist/generate-speech.d.ts.map +1 -1
- package/dist/generate-speech.js +22 -72
- package/dist/generate-speech.js.map +1 -1
- package/dist/index.d.ts +2 -1
- package/dist/index.d.ts.map +1 -1
- package/dist/index.js +1 -1
- package/dist/index.js.map +1 -1
- package/dist/pronunciations/inverse-align.d.ts.map +1 -1
- package/dist/pronunciations/inverse-align.js +6 -1
- package/dist/pronunciations/inverse-align.js.map +1 -1
- package/dist/provider-utils.d.ts +8 -2
- package/dist/provider-utils.d.ts.map +1 -1
- package/dist/provider-utils.js +69 -19
- package/dist/provider-utils.js.map +1 -1
- package/dist/providers/cartesia/index.d.ts.map +1 -1
- package/dist/providers/cartesia/index.js +16 -4
- package/dist/providers/cartesia/index.js.map +1 -1
- package/dist/providers/deepgram/index.d.ts.map +1 -1
- package/dist/providers/deepgram/index.js +10 -2
- package/dist/providers/deepgram/index.js.map +1 -1
- package/dist/providers/elevenlabs/alignment.d.ts.map +1 -1
- package/dist/providers/elevenlabs/alignment.js +57 -23
- package/dist/providers/elevenlabs/alignment.js.map +1 -1
- package/dist/providers/elevenlabs/forced-alignment.d.ts +23 -0
- package/dist/providers/elevenlabs/forced-alignment.d.ts.map +1 -0
- package/dist/providers/elevenlabs/forced-alignment.js +64 -0
- package/dist/providers/elevenlabs/forced-alignment.js.map +1 -0
- package/dist/providers/elevenlabs/index.d.ts +4 -1
- package/dist/providers/elevenlabs/index.d.ts.map +1 -1
- package/dist/providers/elevenlabs/index.js +28 -13
- package/dist/providers/elevenlabs/index.js.map +1 -1
- package/dist/providers/fal/index.d.ts.map +1 -1
- package/dist/providers/fal/index.js +12 -9
- package/dist/providers/fal/index.js.map +1 -1
- package/dist/providers/fish-audio/index.d.ts.map +1 -1
- package/dist/providers/fish-audio/index.js +17 -5
- package/dist/providers/fish-audio/index.js.map +1 -1
- package/dist/providers/gateway/index.d.ts.map +1 -1
- package/dist/providers/gateway/index.js +43 -13
- package/dist/providers/gateway/index.js.map +1 -1
- package/dist/providers/google/index.d.ts.map +1 -1
- package/dist/providers/google/index.js +15 -3
- package/dist/providers/google/index.js.map +1 -1
- package/dist/providers/gradium/index.d.ts.map +1 -1
- package/dist/providers/gradium/index.js +6 -2
- package/dist/providers/gradium/index.js.map +1 -1
- package/dist/providers/hume/index.d.ts.map +1 -1
- package/dist/providers/hume/index.js +22 -6
- package/dist/providers/hume/index.js.map +1 -1
- package/dist/providers/inworld/index.d.ts.map +1 -1
- package/dist/providers/inworld/index.js +13 -5
- package/dist/providers/inworld/index.js.map +1 -1
- package/dist/providers/minimax/index.d.ts.map +1 -1
- package/dist/providers/minimax/index.js +8 -4
- package/dist/providers/minimax/index.js.map +1 -1
- package/dist/providers/mistral/index.d.ts.map +1 -1
- package/dist/providers/mistral/index.js +11 -3
- package/dist/providers/mistral/index.js.map +1 -1
- package/dist/providers/murf/index.d.ts.map +1 -1
- package/dist/providers/murf/index.js +10 -2
- package/dist/providers/murf/index.js.map +1 -1
- package/dist/providers/openai/index.d.ts.map +1 -1
- package/dist/providers/openai/index.js +15 -3
- package/dist/providers/openai/index.js.map +1 -1
- package/dist/providers/resemble/index.d.ts.map +1 -1
- package/dist/providers/resemble/index.js +12 -4
- package/dist/providers/resemble/index.js.map +1 -1
- package/dist/providers/smallest-ai/index.d.ts.map +1 -1
- package/dist/providers/smallest-ai/index.js +6 -2
- package/dist/providers/smallest-ai/index.js.map +1 -1
- package/dist/providers/speechify/index.d.ts.map +1 -1
- package/dist/providers/speechify/index.js +5 -1
- package/dist/providers/speechify/index.js.map +1 -1
- package/dist/providers/xai/index.d.ts.map +1 -1
- package/dist/providers/xai/index.js +11 -3
- package/dist/providers/xai/index.js.map +1 -1
- package/dist/speech-result.d.ts +6 -0
- package/dist/speech-result.d.ts.map +1 -1
- package/dist/speech-result.js.map +1 -1
- package/dist/timestamp-alignment.d.ts +34 -0
- package/dist/timestamp-alignment.d.ts.map +1 -0
- package/dist/timestamp-alignment.js +125 -0
- package/dist/timestamp-alignment.js.map +1 -0
- package/dist/timestamp-finalization.d.ts +15 -0
- package/dist/timestamp-finalization.d.ts.map +1 -0
- package/dist/timestamp-finalization.js +111 -0
- package/dist/timestamp-finalization.js.map +1 -0
- package/dist/timestamp-provider.d.ts +10 -0
- package/dist/timestamp-provider.d.ts.map +1 -0
- package/dist/timestamp-provider.js +2 -0
- package/dist/timestamp-provider.js.map +1 -0
- package/dist/types.d.ts +4 -1
- package/dist/types.d.ts.map +1 -1
- package/package.json +1 -1
- package/dist/default-stt-fallback.d.ts +0 -3
- package/dist/default-stt-fallback.d.ts.map +0 -1
- package/dist/default-stt-fallback.js +0 -11
- package/dist/default-stt-fallback.js.map +0 -1
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/providers/xai/index.ts"],"names":[],"mappings":"AACA,OAAO,EACL,oBAAoB,EACpB,gBAAgB,EAChB,oBAAoB,GACrB,MAAM,sBAAsB,CAAC;AAC9B,OAAO,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,EACL,mBAAmB,EACnB,aAAa,EACb,cAAc,GACf,MAAM,yBAAyB,CAAC;AACjC,OAAO,EAKL,iBAAiB,GAElB,MAAM,0BAA0B,CAAC;AAUlC,MAAM,CAAC,MAAM,eAAe,GAAG,KAAc,CAAC;AAE9C,MAAM,gBAAgB,GAAG;IACvB,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM;CACpC,CAAC;AAEX,iGAAiG;AACjG,MAAM,aAAa,GAAG;IACpB,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;CACI,CAAC;AAEX,MAAM,CAAC,MAAM,UAAU,GAAyB;IAC9C;QACE,EAAE,EAAE,UAAU;QACd,WAAW,EAAE,YAAY;QACzB,SAAS,EAAE,aAAa;QACxB,QAAQ,EAAE,CAAC,WAAW,EAAE,YAAY,EAAE,eAAe,CAAC;QACtD,aAAa,EAAE,MAAM;KACtB;CACO,CAAC;AAEX,MAAM,OAAO,iBAAiB;IACnB,EAAE,GAAG,eAAe,CAAC;IACrB,YAAY,GAAG,UAAU,CAAC;IAE1B,MAAM,GAAG,UAAU,CAAC;IAEZ,MAAM,CAAqB;IAC3B,OAAO,CAAS;IAChB,OAAO,CAA0B;IAElD,YAAY,MAA+B;QACzC,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,OAAO,IAAI,qBAAqB,CAAC;QACvD,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,KAAK,IAAI,UAAU,CAAC,KAAK,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC;IACnE,CAAC;IAED,sFAAsF;IACtF,gBAAgB,CAAC,IAAY;QAC3B,OAAO,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC;IAChC,CAAC;IAEO,SAAS,CAAC,OAIjB;QACC,8FAA8F;QAC9F,MAAM,IAAI,GAA4B;YACpC,QAAQ,EAAE,MAAM;YAChB,GAAG,OAAO,CAAC,eAAe;YAC1B,IAAI,EAAE,OAAO,CAAC,IAAI;SACnB,CAAC;QACF,IAAI,OAAO,CAAC,KAAK,IAAI,IAAI,EAAE,CAAC;YAC1B,IAAI,CAAC,QAAQ,GAAG,OAAO,CAAC,KAAK,CAAC;QAChC,CAAC;QACD,OAAO,IAAI,CAAC;IACd,CAAC;IAEO,gBAAgB,CAAC,IAA6B;QACpD,MAAM,MAAM,GAAG,IAAI,CAAC,aAEP,CAAC;QACd,MAAM,KAAK,GAAG,MAAM,EAAE,KAAK,CAAC;QAC5B,IAAI,KAAK,KAAK,KAAK,EAAE,CAAC;YACpB,OAAO,WAAW,CAAC;QACrB,CAAC;QACD,IAAI,KAAK,KAAK,KAAK,EAAE,CAAC;YACpB,MAAM,IAAI,GACR,OAAO,MAAM,EAAE,WAAW,KAAK,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC,WAAW,CAAC,CAAC,CAAC,IAAI,CAAC;YACtE,OAAO,IAAI,IAAI,IAAI,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC,kBAAkB,IAAI,EAAE,CAAC;QAC/D,CAAC;QACD,IAAI,KAAK,KAAK,OAAO,EAAE,CAAC;YACtB,OAAO,aAAa,CAAC;QACvB,CAAC;QACD,IAAI,KAAK,KAAK,MAAM,EAAE,CAAC;YACrB,OAAO,YAAY,CAAC;QACtB,CAAC;QACD,OAAO,YAAY,CAAC;IACtB,CAAC;IAED,KAAK,CAAC,QAAQ,CAAC,OAOd;QAIC,MAAM,IAAI,GAAG,IAAI,CAAC,SAAS,CAAC,OAAO,CAAC,CAAC;QACrC,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,IAAI,CAAC,OAAO,MAAM,EAAE;YACzD,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,cAAc,EAAE,kBAAkB;gBAClC,aAAa,EAAE,UAAU,aAAa,CAAC,IAAI,CAAC,MAAM,EAAE,aAAa,EAAE,KAAK,CAAC,EAAE;gBAC3E,cAAc,EAAE,cAAc;gBAC9B,GAAG,OAAO,CAAC,OAAO;aACnB;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC;YAC1B,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,MAAM,mBAAmB,CAAC,QAAQ,CAAC,CAAC;
|
|
1
|
+
{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/providers/xai/index.ts"],"names":[],"mappings":"AACA,OAAO,EACL,oBAAoB,EACpB,gBAAgB,EAChB,oBAAoB,GACrB,MAAM,sBAAsB,CAAC;AAC9B,OAAO,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,EACL,mBAAmB,EACnB,aAAa,EACb,cAAc,GACf,MAAM,yBAAyB,CAAC;AACjC,OAAO,EAKL,iBAAiB,GAElB,MAAM,0BAA0B,CAAC;AAUlC,MAAM,CAAC,MAAM,eAAe,GAAG,KAAc,CAAC;AAE9C,MAAM,gBAAgB,GAAG;IACvB,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,MAAM;CACpC,CAAC;AAEX,iGAAiG;AACjG,MAAM,aAAa,GAAG;IACpB,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;CACI,CAAC;AAEX,MAAM,CAAC,MAAM,UAAU,GAAyB;IAC9C;QACE,EAAE,EAAE,UAAU;QACd,WAAW,EAAE,YAAY;QACzB,SAAS,EAAE,aAAa;QACxB,QAAQ,EAAE,CAAC,WAAW,EAAE,YAAY,EAAE,eAAe,CAAC;QACtD,aAAa,EAAE,MAAM;KACtB;CACO,CAAC;AAEX,MAAM,OAAO,iBAAiB;IACnB,EAAE,GAAG,eAAe,CAAC;IACrB,YAAY,GAAG,UAAU,CAAC;IAE1B,MAAM,GAAG,UAAU,CAAC;IAEZ,MAAM,CAAqB;IAC3B,OAAO,CAAS;IAChB,OAAO,CAA0B;IAElD,YAAY,MAA+B;QACzC,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,OAAO,IAAI,qBAAqB,CAAC;QACvD,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC,KAAK,IAAI,UAAU,CAAC,KAAK,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC;IACnE,CAAC;IAED,sFAAsF;IACtF,gBAAgB,CAAC,IAAY;QAC3B,OAAO,EAAE,IAAI,EAAE,QAAQ,EAAE,EAAE,EAAE,CAAC;IAChC,CAAC;IAEO,SAAS,CAAC,OAIjB;QACC,8FAA8F;QAC9F,MAAM,IAAI,GAA4B;YACpC,QAAQ,EAAE,MAAM;YAChB,GAAG,OAAO,CAAC,eAAe;YAC1B,IAAI,EAAE,OAAO,CAAC,IAAI;SACnB,CAAC;QACF,IAAI,OAAO,CAAC,KAAK,IAAI,IAAI,EAAE,CAAC;YAC1B,IAAI,CAAC,QAAQ,GAAG,OAAO,CAAC,KAAK,CAAC;QAChC,CAAC;QACD,OAAO,IAAI,CAAC;IACd,CAAC;IAEO,gBAAgB,CAAC,IAA6B;QACpD,MAAM,MAAM,GAAG,IAAI,CAAC,aAEP,CAAC;QACd,MAAM,KAAK,GAAG,MAAM,EAAE,KAAK,CAAC;QAC5B,IAAI,KAAK,KAAK,KAAK,EAAE,CAAC;YACpB,OAAO,WAAW,CAAC;QACrB,CAAC;QACD,IAAI,KAAK,KAAK,KAAK,EAAE,CAAC;YACpB,MAAM,IAAI,GACR,OAAO,MAAM,EAAE,WAAW,KAAK,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC,WAAW,CAAC,CAAC,CAAC,IAAI,CAAC;YACtE,OAAO,IAAI,IAAI,IAAI,CAAC,CAAC,CAAC,WAAW,CAAC,CAAC,CAAC,kBAAkB,IAAI,EAAE,CAAC;QAC/D,CAAC;QACD,IAAI,KAAK,KAAK,OAAO,EAAE,CAAC;YACtB,OAAO,aAAa,CAAC;QACvB,CAAC;QACD,IAAI,KAAK,KAAK,MAAM,EAAE,CAAC;YACrB,OAAO,YAAY,CAAC;QACtB,CAAC;QACD,OAAO,YAAY,CAAC;IACtB,CAAC;IAED,KAAK,CAAC,QAAQ,CAAC,OAOd;QAIC,MAAM,IAAI,GAAG,IAAI,CAAC,SAAS,CAAC,OAAO,CAAC,CAAC;QACrC,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,IAAI,CAAC,OAAO,MAAM,EAAE;YACzD,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,cAAc,EAAE,kBAAkB;gBAClC,aAAa,EAAE,UAAU,aAAa,CAAC,IAAI,CAAC,MAAM,EAAE,aAAa,EAAE,KAAK,CAAC,EAAE;gBAC3E,cAAc,EAAE,cAAc;gBAC9B,GAAG,OAAO,CAAC,OAAO;aACnB;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC;YAC1B,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,MAAM,mBAAmB,CAAC,QAAQ,EAAE;YAClC,QAAQ,EAAE,IAAI,CAAC,EAAE;YACjB,KAAK,EAAE,OAAO,CAAC,OAAO;YACtB,KAAK,EAAE,WAAW;SACnB,CAAC,CAAC;QAEH,MAAM,WAAW,GAAG,MAAM,QAAQ,CAAC,WAAW,EAAE,CAAC;QACjD,yHAAyH;QACzH,MAAM,SAAS,GAAG,IAAI,CAAC,gBAAgB,CAAC,IAAI,CAAC,CAAC;QAE9C,OAAO;YACL,KAAK,EAAE,IAAI,UAAU,CAAC,WAAW,CAAC;YAClC,SAAS;SACV,CAAC;IACJ,CAAC;IAED,KAAK,CAAC,MAAM,CAAC,OAOZ;QAIC,MAAM,IAAI,GAAG,IAAI,CAAC,SAAS,CAAC,OAAO,CAAC,CAAC;QACrC,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,IAAI,CAAC,OAAO,MAAM,EAAE;YACzD,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,cAAc,EAAE,kBAAkB;gBAClC,aAAa,EAAE,UAAU,aAAa,CAAC,IAAI,CAAC,MAAM,EAAE,aAAa,EAAE,KAAK,CAAC,EAAE;gBAC3E,cAAc,EAAE,cAAc;gBAC9B,GAAG,OAAO,CAAC,OAAO;aACnB;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC;YAC1B,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,MAAM,mBAAmB,CAAC,QAAQ,EAAE;YAClC,QAAQ,EAAE,IAAI,CAAC,EAAE;YACjB,KAAK,EAAE,OAAO,CAAC,OAAO;YACtB,KAAK,EAAE,WAAW;SACnB,CAAC,CAAC;QAEH,IAAI,CAAC,QAAQ,CAAC,IAAI,EAAE,CAAC;YACnB,MAAM,IAAI,KAAK,CAAC,OAAO,OAAO,CAAC,OAAO,wBAAwB,CAAC,CAAC;QAClE,CAAC;QAED,OAAO;YACL,MAAM,EAAE,QAAQ,CAAC,IAAI;YACrB,SAAS,EAAE,IAAI,CAAC,gBAAgB,CAAC,IAAI,CAAC;SACvC,CAAC;IACJ,CAAC;IAED,oBAAoB,CAAC,OAAe;QAClC,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,OAAO,CAAC,EAAE,CAAC;YAC/C,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,OAAO,gBAAgB,CAAC;IAC1B,CAAC;IAED,gBAAgB,CAAC,OAAe,EAAE,IAA8B;QAC9D,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,OAAO,CAAC,EAAE,CAAC;YAC/C,OAAO;QACT,CAAC;QACD,MAAM,IAAI,GAAG,iBAAiB,CAC5B,OAAO,OAAO,EAAE,EAChB,IAAI,CAAC,oBAAoB,CAAC,OAAO,CAAC,EAClC,IAAI,EAAE,UAAU,CACjB,CAAC;QACF,OAAO;YACL,eAAe,EAAE,EAAE,aAAa,EAAE,EAAE,KAAK,EAAE,KAAK,EAAE,WAAW,EAAE,IAAI,EAAE,EAAE;YACvE,SAAS,EAAE,WAAW;SACvB,CAAC;IACJ,CAAC;IAED,KAAK,CAAC,UAAU,CACd,OAAkC;QAElC,MAAM,QAAQ,GAAa,EAAE,CAAC;QAC9B,MAAM,QAAQ,GAAG,oBAAoB,CAAC,KAAK,EAAE,OAAO,CAAC,QAAQ,EAAE,QAAQ,CAAC,CAAC;QAEzE,MAAM,IAAI,GAAG,IAAI,QAAQ,EAAE,CAAC;QAC5B,IAAI,CAAC,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,IAAI,CAAC,CAAC;QAClC,IAAI,CAAC,MAAM,CAAC,UAAU,EAAE,QAAQ,CAAC,CAAC;QAClC,MAAM,MAAM,GAAG,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC;QAClC,gBAAgB,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC;QAE1C,KAAK,MAAM,CAAC,GAAG,EAAE,KAAK,CAAC,IAAI,MAAM,CAAC,OAAO,CAAC,OAAO,CAAC,eAAe,IAAI,EAAE,CAAC,EAAE,CAAC;YACzE,oBAAoB,CAAC,IAAI,EAAE,GAAG,EAAE,KAAK,CAAC,CAAC;QACzC,CAAC;QAED,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,IAAI,CAAC,OAAO,gBAAgB,EAAE;YACnE,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACP,aAAa,EAAE,UAAU,aAAa,CAAC,IAAI,CAAC,MAAM,EAAE,aAAa,EAAE,KAAK,CAAC,EAAE;gBAC3E,cAAc,EAAE,cAAc;gBAC9B,GAAG,OAAO,CAAC,OAAO;aACnB;YACD,IAAI,EAAE,IAAI;YACV,MAAM,EAAE,OAAO,CAAC,WAAW;SAC5B,CAAC,CAAC;QAEH,MAAM,mBAAmB,CAAC,QAAQ,EAAE,EAAE,QAAQ,EAAE,IAAI,CAAC,EAAE,EAAE,CAAC,CAAC;QAE3D,MAAM,IAAI,GAAG,CAAC,MAAM,QAAQ,CAAC,IAAI,EAAE,CAA4B,CAAC;QAChE,MAAM,OAAO,GAAG,IAAI,CAAC,QAAQ,CAAC;QAC9B,IAAI,OAAO,OAAO,KAAK,QAAQ,EAAE,CAAC;YAChC,MAAM,IAAI,cAAc,CAAC,sCAAsC,CAAC,CAAC;QACnE,CAAC;QACD,OAAO;YACL,OAAO;YACP,GAAG,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,QAAQ,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;YACxC,gBAAgB,EAAE,IAAI;SACvB,CAAC;IACJ,CAAC;IAED,mBAAmB,CAAC,OAAe,EAAE,MAAmB;QACtD,IAAI,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,OAAO,CAAC,EAAE,CAAC;YAC/C,OAAO;QACT,CAAC;QACD,QAAQ,MAAM,CAAC,MAAM,EAAE,CAAC;YACtB,KAAK,KAAK,CAAC,CAAC,CAAC;gBACX,MAAM,IAAI,GAAG,iBAAiB,CAC5B,OAAO,OAAO,EAAE,EAChB,IAAI,CAAC,oBAAoB,CAAC,OAAO,CAAC,EAClC,MAAM,CAAC,UAAU,CAClB,CAAC;gBACF,OAAO;oBACL,eAAe,EAAE;wBACf,aAAa,EAAE,EAAE,KAAK,EAAE,KAAK,EAAE,WAAW,EAAE,IAAI,EAAE;qBACnD;oBACD,iBAAiB,EAAE,WAAW;iBAC/B,CAAC;YACJ,CAAC;YACD,KAAK,KAAK;gBACR,OAAO;oBACL,eAAe,EAAE,EAAE,aAAa,EAAE,EAAE,KAAK,EAAE,KAAK,EAAE,EAAE;oBACpD,iBAAiB,EAAE,YAAY;iBAChC,CAAC;YACJ,KAAK,KAAK,CAAC,CAAC,CAAC;gBACX,MAAM,IAAI,GAAG,iBAAiB,CAC5B,OAAO,OAAO,EAAE,EAChB,IAAI,CAAC,oBAAoB,CAAC,OAAO,CAAC,EAClC,MAAM,CAAC,UAAU,CAClB,CAAC;gBACF,OAAO;oBACL,eAAe,EAAE;wBACf,aAAa,EAAE,EAAE,KAAK,EAAE,KAAK,EAAE,WAAW,EAAE,IAAI,EAAE;qBACnD;oBACD,iBAAiB,EAAE,kBAAkB,IAAI,EAAE;iBAC5C,CAAC;YACJ,CAAC;YACD;gBACE,OAAO;QACX,CAAC;IACH,CAAC;CACF;AAED,MAAM,UAAU,SAAS,CAAC,SAAkC,EAAE;IAC5D,MAAM,QAAQ,GAAG,IAAI,iBAAiB,CAAC,MAAM,CAAC,CAAC;IAC/C,MAAM,WAAW,GAAG,MAAM,CAAC,WAAW,CAAC;IACvC,OAAO,SAAS,GAAG,CAAC,OAAgB;QAClC,OAAO;YACL,QAAQ;YACR,OAAO,EAAE,OAAO,IAAI,QAAQ,CAAC,YAAY;YACzC,GAAG,CAAC,WAAW,IAAI,EAAE,WAAW,EAAE,CAAC;SACpC,CAAC;IACJ,CAAC,CAAC;AACJ,CAAC"}
|
package/dist/speech-result.d.ts
CHANGED
|
@@ -12,6 +12,9 @@ export interface SpeechResult {
|
|
|
12
12
|
readonly timestamps?: readonly WordTimestamp[];
|
|
13
13
|
readonly warnings?: string[];
|
|
14
14
|
}
|
|
15
|
+
export interface SpeechResultWithTimestamps extends SpeechResult {
|
|
16
|
+
readonly timestamps: readonly WordTimestamp[];
|
|
17
|
+
}
|
|
15
18
|
export interface ConversationMetadata extends SpeechMetadata {
|
|
16
19
|
readonly perTurn?: readonly SpeechMetadata[];
|
|
17
20
|
}
|
|
@@ -19,6 +22,9 @@ export interface ConversationResult extends Omit<SpeechResult, "metadata" | "tim
|
|
|
19
22
|
readonly metadata: ConversationMetadata;
|
|
20
23
|
readonly timestamps?: readonly ConversationWordTimestamp[];
|
|
21
24
|
}
|
|
25
|
+
export interface ConversationResultWithTimestamps extends ConversationResult {
|
|
26
|
+
readonly timestamps: readonly ConversationWordTimestamp[];
|
|
27
|
+
}
|
|
22
28
|
export declare class DefaultGeneratedAudioFile implements GeneratedAudioFile {
|
|
23
29
|
readonly mediaType: string;
|
|
24
30
|
private readonly _data;
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"speech-result.d.ts","sourceRoot":"","sources":["../src/speech-result.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,eAAe,CAAC;AACpD,OAAO,KAAK,EAAE,yBAAyB,EAAE,aAAa,EAAE,MAAM,iBAAiB,CAAC;AAEhF,MAAM,WAAW,kBAAkB;IACjC,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;IACxB,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;IAC3B,QAAQ,CAAC,UAAU,EAAE,UAAU,CAAC;CACjC;AAED,MAAM,WAAW,YAAY;IAC3B,QAAQ,CAAC,KAAK,EAAE,kBAAkB,CAAC;IACnC,QAAQ,CAAC,QAAQ,EAAE,cAAc,CAAC;IAClC,QAAQ,CAAC,gBAAgB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IACpD,QAAQ,CAAC,UAAU,CAAC,EAAE,SAAS,aAAa,EAAE,CAAC;IAC/C,QAAQ,CAAC,QAAQ,CAAC,EAAE,MAAM,EAAE,CAAC;CAC9B;AAED,MAAM,WAAW,oBAAqB,SAAQ,cAAc;IAG1D,QAAQ,CAAC,OAAO,CAAC,EAAE,SAAS,cAAc,EAAE,CAAC;CAC9C;AAED,MAAM,WAAW,kBACf,SAAQ,IAAI,CAAC,YAAY,EAAE,UAAU,GAAG,YAAY,CAAC;IACrD,QAAQ,CAAC,QAAQ,EAAE,oBAAoB,CAAC;IACxC,QAAQ,CAAC,UAAU,CAAC,EAAE,SAAS,yBAAyB,EAAE,CAAC;CAC5D;AAED,qBAAa,yBAA0B,YAAW,kBAAkB;IAClE,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;IAE3B,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAsB;IAC5C,OAAO,CAAC,WAAW,CAAC,CAAa;IACjC,OAAO,CAAC,OAAO,CAAC,CAAS;gBAEb,EACV,IAAI,EACJ,SAAS,GACV,EAAE;QAAE,IAAI,EAAE,MAAM,GAAG,UAAU,CAAC;QAAC,SAAS,EAAE,MAAM,CAAA;KAAE;IAKnD,IAAI,UAAU,IAAI,UAAU,CAe3B;IAED,IAAI,MAAM,IAAI,MAAM,CAcnB;CACF"}
|
|
1
|
+
{"version":3,"file":"speech-result.d.ts","sourceRoot":"","sources":["../src/speech-result.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,eAAe,CAAC;AACpD,OAAO,KAAK,EAAE,yBAAyB,EAAE,aAAa,EAAE,MAAM,iBAAiB,CAAC;AAEhF,MAAM,WAAW,kBAAkB;IACjC,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;IACxB,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;IAC3B,QAAQ,CAAC,UAAU,EAAE,UAAU,CAAC;CACjC;AAED,MAAM,WAAW,YAAY;IAC3B,QAAQ,CAAC,KAAK,EAAE,kBAAkB,CAAC;IACnC,QAAQ,CAAC,QAAQ,EAAE,cAAc,CAAC;IAClC,QAAQ,CAAC,gBAAgB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IACpD,QAAQ,CAAC,UAAU,CAAC,EAAE,SAAS,aAAa,EAAE,CAAC;IAC/C,QAAQ,CAAC,QAAQ,CAAC,EAAE,MAAM,EAAE,CAAC;CAC9B;AAED,MAAM,WAAW,0BAA2B,SAAQ,YAAY;IAC9D,QAAQ,CAAC,UAAU,EAAE,SAAS,aAAa,EAAE,CAAC;CAC/C;AAED,MAAM,WAAW,oBAAqB,SAAQ,cAAc;IAG1D,QAAQ,CAAC,OAAO,CAAC,EAAE,SAAS,cAAc,EAAE,CAAC;CAC9C;AAED,MAAM,WAAW,kBACf,SAAQ,IAAI,CAAC,YAAY,EAAE,UAAU,GAAG,YAAY,CAAC;IACrD,QAAQ,CAAC,QAAQ,EAAE,oBAAoB,CAAC;IACxC,QAAQ,CAAC,UAAU,CAAC,EAAE,SAAS,yBAAyB,EAAE,CAAC;CAC5D;AAED,MAAM,WAAW,gCAAiC,SAAQ,kBAAkB;IAC1E,QAAQ,CAAC,UAAU,EAAE,SAAS,yBAAyB,EAAE,CAAC;CAC3D;AAED,qBAAa,yBAA0B,YAAW,kBAAkB;IAClE,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;IAE3B,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAsB;IAC5C,OAAO,CAAC,WAAW,CAAC,CAAa;IACjC,OAAO,CAAC,OAAO,CAAC,CAAS;gBAEb,EACV,IAAI,EACJ,SAAS,GACV,EAAE;QAAE,IAAI,EAAE,MAAM,GAAG,UAAU,CAAC;QAAC,SAAS,EAAE,MAAM,CAAA;KAAE;IAKnD,IAAI,UAAU,IAAI,UAAU,CAe3B;IAED,IAAI,MAAM,IAAI,MAAM,CAcnB;CACF"}
|
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"speech-result.js","sourceRoot":"","sources":["../src/speech-result.ts"],"names":[],"mappings":"
|
|
1
|
+
{"version":3,"file":"speech-result.js","sourceRoot":"","sources":["../src/speech-result.ts"],"names":[],"mappings":"AAqCA,MAAM,OAAO,yBAAyB;IAC3B,SAAS,CAAS;IAEV,KAAK,CAAsB;IACpC,WAAW,CAAc;IACzB,OAAO,CAAU;IAEzB,YAAY,EACV,IAAI,EACJ,SAAS,GACwC;QACjD,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC;QAClB,IAAI,CAAC,SAAS,GAAG,SAAS,CAAC;IAC7B,CAAC;IAED,IAAI,UAAU;QACZ,IAAI,IAAI,CAAC,WAAW,IAAI,IAAI,EAAE,CAAC;YAC7B,OAAO,IAAI,CAAC,WAAW,CAAC;QAC1B,CAAC;QACD,IAAI,IAAI,CAAC,KAAK,YAAY,UAAU,EAAE,CAAC;YACrC,IAAI,CAAC,WAAW,GAAG,IAAI,CAAC,KAAK,CAAC;QAChC,CAAC;aAAM,CAAC;YACN,MAAM,YAAY,GAAG,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;YACtC,MAAM,KAAK,GAAG,IAAI,UAAU,CAAC,YAAY,CAAC,MAAM,CAAC,CAAC;YAClD,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,YAAY,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;gBAC7C,KAAK,CAAC,CAAC,CAAC,GAAG,YAAY,CAAC,UAAU,CAAC,CAAC,CAAC,CAAC;YACxC,CAAC;YACD,IAAI,CAAC,WAAW,GAAG,KAAK,CAAC;QAC3B,CAAC;QACD,OAAO,IAAI,CAAC,WAAW,CAAC;IAC1B,CAAC;IAED,IAAI,MAAM;QACR,IAAI,IAAI,CAAC,OAAO,IAAI,IAAI,EAAE,CAAC;YACzB,OAAO,IAAI,CAAC,OAAO,CAAC;QACtB,CAAC;QACD,IAAI,OAAO,IAAI,CAAC,KAAK,KAAK,QAAQ,EAAE,CAAC;YACnC,IAAI,CAAC,OAAO,GAAG,IAAI,CAAC,KAAK,CAAC;QAC5B,CAAC;aAAM,CAAC;YACN,IAAI,YAAY,GAAG,EAAE,CAAC;YACtB,KAAK,MAAM,IAAI,IAAI,IAAI,CAAC,KAAK,EAAE,CAAC;gBAC9B,YAAY,IAAI,MAAM,CAAC,YAAY,CAAC,IAAI,CAAC,CAAC;YAC5C,CAAC;YACD,IAAI,CAAC,OAAO,GAAG,IAAI,CAAC,YAAY,CAAC,CAAC;QACpC,CAAC;QACD,OAAO,IAAI,CAAC,OAAO,CAAC;IACtB,CAAC;CACF"}
|
|
@@ -0,0 +1,34 @@
|
|
|
1
|
+
import type { Edit } from "./pronunciations/types.js";
|
|
2
|
+
import { type ResolvedModel } from "./speech-provider.js";
|
|
3
|
+
import type { TimestampProvider } from "./timestamp-provider.js";
|
|
4
|
+
import type { WordTimestamp } from "./timestamps.js";
|
|
5
|
+
export interface TimestampAlignmentResult {
|
|
6
|
+
readonly timestamps?: readonly WordTimestamp[];
|
|
7
|
+
}
|
|
8
|
+
export interface TimestampAlignmentPlan {
|
|
9
|
+
readonly includeNative: boolean;
|
|
10
|
+
resolve(args: {
|
|
11
|
+
readonly abortSignal?: AbortSignal;
|
|
12
|
+
readonly audio: Uint8Array;
|
|
13
|
+
readonly audioDurationMs?: number;
|
|
14
|
+
readonly mediaType: string;
|
|
15
|
+
readonly originalText: string;
|
|
16
|
+
readonly pronunciationEdits: readonly Edit[];
|
|
17
|
+
readonly providerTimestamps?: readonly WordTimestamp[];
|
|
18
|
+
readonly synthesizedText: string;
|
|
19
|
+
}): Promise<TimestampAlignmentResult>;
|
|
20
|
+
}
|
|
21
|
+
export declare function deriveTimestampsViaProvider(args: {
|
|
22
|
+
readonly abortSignal?: AbortSignal;
|
|
23
|
+
readonly audio: Uint8Array;
|
|
24
|
+
readonly mediaType: string;
|
|
25
|
+
readonly provider: TimestampProvider;
|
|
26
|
+
readonly text: string;
|
|
27
|
+
}): Promise<readonly WordTimestamp[]>;
|
|
28
|
+
export declare function prepareTimestampAlignment(args: {
|
|
29
|
+
readonly modelIdentifier: string;
|
|
30
|
+
readonly request: boolean;
|
|
31
|
+
readonly resolved: ResolvedModel;
|
|
32
|
+
readonly timestampProvider?: TimestampProvider;
|
|
33
|
+
}): TimestampAlignmentPlan;
|
|
34
|
+
//# sourceMappingURL=timestamp-alignment.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"timestamp-alignment.d.ts","sourceRoot":"","sources":["../src/timestamp-alignment.ts"],"names":[],"mappings":"AAQA,OAAO,KAAK,EAAE,IAAI,EAAE,MAAM,2BAA2B,CAAC;AACtD,OAAO,EAGL,KAAK,aAAa,EACnB,MAAM,sBAAsB,CAAC;AAE9B,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,yBAAyB,CAAC;AACjE,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,iBAAiB,CAAC;AAErD,MAAM,WAAW,wBAAwB;IACvC,QAAQ,CAAC,UAAU,CAAC,EAAE,SAAS,aAAa,EAAE,CAAC;CAChD;AAED,MAAM,WAAW,sBAAsB;IACrC,QAAQ,CAAC,aAAa,EAAE,OAAO,CAAC;IAChC,OAAO,CAAC,IAAI,EAAE;QACZ,QAAQ,CAAC,WAAW,CAAC,EAAE,WAAW,CAAC;QACnC,QAAQ,CAAC,KAAK,EAAE,UAAU,CAAC;QAC3B,QAAQ,CAAC,eAAe,CAAC,EAAE,MAAM,CAAC;QAClC,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;QAC3B,QAAQ,CAAC,YAAY,EAAE,MAAM,CAAC;QAC9B,QAAQ,CAAC,kBAAkB,EAAE,SAAS,IAAI,EAAE,CAAC;QAC7C,QAAQ,CAAC,kBAAkB,CAAC,EAAE,SAAS,aAAa,EAAE,CAAC;QACvD,QAAQ,CAAC,eAAe,EAAE,MAAM,CAAC;KAClC,GAAG,OAAO,CAAC,wBAAwB,CAAC,CAAC;CACvC;AAwBD,wBAAsB,2BAA2B,CAAC,IAAI,EAAE;IACtD,QAAQ,CAAC,WAAW,CAAC,EAAE,WAAW,CAAC;IACnC,QAAQ,CAAC,KAAK,EAAE,UAAU,CAAC;IAC3B,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;IAC3B,QAAQ,CAAC,QAAQ,EAAE,iBAAiB,CAAC;IACrC,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;CACvB,GAAG,OAAO,CAAC,SAAS,aAAa,EAAE,CAAC,CAWpC;AA0BD,wBAAgB,yBAAyB,CAAC,IAAI,EAAE;IAC9C,QAAQ,CAAC,eAAe,EAAE,MAAM,CAAC;IACjC,QAAQ,CAAC,OAAO,EAAE,OAAO,CAAC;IAC1B,QAAQ,CAAC,QAAQ,EAAE,aAAa,CAAC;IACjC,QAAQ,CAAC,iBAAiB,CAAC,EAAE,iBAAiB,CAAC;CAChD,GAAG,sBAAsB,CAiFzB"}
|
|
@@ -0,0 +1,125 @@
|
|
|
1
|
+
import { deriveTimestampsViaSTT } from "./derive-timestamps.js";
|
|
2
|
+
import { TimestampProviderRequiredError, TimestampValidationError, withProviderErrorStage, } from "./errors.js";
|
|
3
|
+
import { debug } from "./logger.js";
|
|
4
|
+
import { inverseAlign } from "./pronunciations/inverse-align.js";
|
|
5
|
+
import { isSpeechGatewayModel, modelDeclaresNativeTimestamps, } from "./speech-provider.js";
|
|
6
|
+
import { finalizeTimestamps } from "./timestamp-finalization.js";
|
|
7
|
+
function finalizeCandidate(args) {
|
|
8
|
+
const finalized = finalizeTimestamps({
|
|
9
|
+
timestamps: args.timestamps,
|
|
10
|
+
text: args.text,
|
|
11
|
+
audioDurationMs: args.audioDurationMs,
|
|
12
|
+
});
|
|
13
|
+
if (!finalized.ok) {
|
|
14
|
+
throw new TimestampValidationError({
|
|
15
|
+
reason: finalized.reason,
|
|
16
|
+
source: args.source,
|
|
17
|
+
});
|
|
18
|
+
}
|
|
19
|
+
debug(`${args.source}: validated ${finalized.timestamps.length} timestamps.`);
|
|
20
|
+
return finalized.timestamps;
|
|
21
|
+
}
|
|
22
|
+
export async function deriveTimestampsViaProvider(args) {
|
|
23
|
+
try {
|
|
24
|
+
return await args.provider.align({
|
|
25
|
+
abortSignal: args.abortSignal,
|
|
26
|
+
audio: args.audio,
|
|
27
|
+
mediaType: args.mediaType,
|
|
28
|
+
text: args.text,
|
|
29
|
+
});
|
|
30
|
+
}
|
|
31
|
+
catch (error) {
|
|
32
|
+
throw withProviderErrorStage(error, "alignment");
|
|
33
|
+
}
|
|
34
|
+
}
|
|
35
|
+
function projectPublicText(args) {
|
|
36
|
+
if (args.pronunciationEdits.length === 0) {
|
|
37
|
+
return args.timestamps;
|
|
38
|
+
}
|
|
39
|
+
const mapped = inverseAlign(args.timestamps, args.substitutedText, args.pronunciationEdits);
|
|
40
|
+
return finalizeCandidate({
|
|
41
|
+
timestamps: mapped,
|
|
42
|
+
text: args.originalText,
|
|
43
|
+
audioDurationMs: args.audioDurationMs,
|
|
44
|
+
source: "pronunciation projection",
|
|
45
|
+
});
|
|
46
|
+
}
|
|
47
|
+
export function prepareTimestampAlignment(args) {
|
|
48
|
+
const isGateway = isSpeechGatewayModel(args.resolved);
|
|
49
|
+
const hasNative = modelDeclaresNativeTimestamps(args.resolved);
|
|
50
|
+
const includeNative = args.request && (isGateway || hasNative);
|
|
51
|
+
if (args.request &&
|
|
52
|
+
!(isGateway ||
|
|
53
|
+
hasNative ||
|
|
54
|
+
args.timestampProvider ||
|
|
55
|
+
args.resolved.fallbackSTT)) {
|
|
56
|
+
throw new TimestampProviderRequiredError(args.modelIdentifier);
|
|
57
|
+
}
|
|
58
|
+
if (!args.request) {
|
|
59
|
+
debug(`${args.modelIdentifier}: timestamps disabled.`);
|
|
60
|
+
}
|
|
61
|
+
else if (includeNative) {
|
|
62
|
+
debug(`${args.modelIdentifier}: requesting native timestamps.`);
|
|
63
|
+
}
|
|
64
|
+
else if (args.timestampProvider) {
|
|
65
|
+
debug(`${args.modelIdentifier}: using the supplied timestamp provider.`);
|
|
66
|
+
}
|
|
67
|
+
else {
|
|
68
|
+
debug(`${args.modelIdentifier}: using the configured legacy STT fallback.`);
|
|
69
|
+
}
|
|
70
|
+
return {
|
|
71
|
+
includeNative,
|
|
72
|
+
resolve: async (resolutionArgs) => {
|
|
73
|
+
if (!args.request) {
|
|
74
|
+
return {};
|
|
75
|
+
}
|
|
76
|
+
let timestamps;
|
|
77
|
+
let source;
|
|
78
|
+
if (includeNative) {
|
|
79
|
+
timestamps = resolutionArgs.providerTimestamps ?? [];
|
|
80
|
+
source = args.modelIdentifier;
|
|
81
|
+
}
|
|
82
|
+
else if (args.timestampProvider) {
|
|
83
|
+
timestamps = await deriveTimestampsViaProvider({
|
|
84
|
+
abortSignal: resolutionArgs.abortSignal,
|
|
85
|
+
audio: resolutionArgs.audio,
|
|
86
|
+
mediaType: resolutionArgs.mediaType,
|
|
87
|
+
provider: args.timestampProvider,
|
|
88
|
+
text: resolutionArgs.synthesizedText,
|
|
89
|
+
});
|
|
90
|
+
source = "timestampProvider";
|
|
91
|
+
}
|
|
92
|
+
else {
|
|
93
|
+
const fallback = args.resolved.fallbackSTT;
|
|
94
|
+
if (!fallback) {
|
|
95
|
+
throw new TimestampProviderRequiredError(args.modelIdentifier);
|
|
96
|
+
}
|
|
97
|
+
timestamps = await deriveTimestampsViaSTT({
|
|
98
|
+
ttsModel: args.modelIdentifier,
|
|
99
|
+
audio: resolutionArgs.audio,
|
|
100
|
+
mediaType: resolutionArgs.mediaType,
|
|
101
|
+
text: resolutionArgs.synthesizedText,
|
|
102
|
+
timestampFallback: fallback,
|
|
103
|
+
abortSignal: resolutionArgs.abortSignal,
|
|
104
|
+
});
|
|
105
|
+
source = `${fallback.provider.id}/${fallback.modelId}`;
|
|
106
|
+
}
|
|
107
|
+
const finalized = finalizeCandidate({
|
|
108
|
+
timestamps,
|
|
109
|
+
text: resolutionArgs.synthesizedText,
|
|
110
|
+
audioDurationMs: resolutionArgs.audioDurationMs,
|
|
111
|
+
source,
|
|
112
|
+
});
|
|
113
|
+
return {
|
|
114
|
+
timestamps: projectPublicText({
|
|
115
|
+
timestamps: finalized,
|
|
116
|
+
audioDurationMs: resolutionArgs.audioDurationMs,
|
|
117
|
+
originalText: resolutionArgs.originalText,
|
|
118
|
+
pronunciationEdits: resolutionArgs.pronunciationEdits,
|
|
119
|
+
substitutedText: resolutionArgs.synthesizedText,
|
|
120
|
+
}),
|
|
121
|
+
};
|
|
122
|
+
},
|
|
123
|
+
};
|
|
124
|
+
}
|
|
125
|
+
//# sourceMappingURL=timestamp-alignment.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"timestamp-alignment.js","sourceRoot":"","sources":["../src/timestamp-alignment.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,sBAAsB,EAAE,MAAM,wBAAwB,CAAC;AAChE,OAAO,EACL,8BAA8B,EAC9B,wBAAwB,EACxB,sBAAsB,GACvB,MAAM,aAAa,CAAC;AACrB,OAAO,EAAE,KAAK,EAAE,MAAM,aAAa,CAAC;AACpC,OAAO,EAAE,YAAY,EAAE,MAAM,mCAAmC,CAAC;AAEjE,OAAO,EACL,oBAAoB,EACpB,6BAA6B,GAE9B,MAAM,sBAAsB,CAAC;AAC9B,OAAO,EAAE,kBAAkB,EAAE,MAAM,6BAA6B,CAAC;AAsBjE,SAAS,iBAAiB,CAAC,IAK1B;IACC,MAAM,SAAS,GAAG,kBAAkB,CAAC;QACnC,UAAU,EAAE,IAAI,CAAC,UAAU;QAC3B,IAAI,EAAE,IAAI,CAAC,IAAI;QACf,eAAe,EAAE,IAAI,CAAC,eAAe;KACtC,CAAC,CAAC;IACH,IAAI,CAAC,SAAS,CAAC,EAAE,EAAE,CAAC;QAClB,MAAM,IAAI,wBAAwB,CAAC;YACjC,MAAM,EAAE,SAAS,CAAC,MAAM;YACxB,MAAM,EAAE,IAAI,CAAC,MAAM;SACpB,CAAC,CAAC;IACL,CAAC;IAED,KAAK,CAAC,GAAG,IAAI,CAAC,MAAM,eAAe,SAAS,CAAC,UAAU,CAAC,MAAM,cAAc,CAAC,CAAC;IAC9E,OAAO,SAAS,CAAC,UAAU,CAAC;AAC9B,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,2BAA2B,CAAC,IAMjD;IACC,IAAI,CAAC;QACH,OAAO,MAAM,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC;YAC/B,WAAW,EAAE,IAAI,CAAC,WAAW;YAC7B,KAAK,EAAE,IAAI,CAAC,KAAK;YACjB,SAAS,EAAE,IAAI,CAAC,SAAS;YACzB,IAAI,EAAE,IAAI,CAAC,IAAI;SAChB,CAAC,CAAC;IACL,CAAC;IAAC,OAAO,KAAK,EAAE,CAAC;QACf,MAAM,sBAAsB,CAAC,KAAK,EAAE,WAAW,CAAC,CAAC;IACnD,CAAC;AACH,CAAC;AAED,SAAS,iBAAiB,CAAC,IAM1B;IACC,IAAI,IAAI,CAAC,kBAAkB,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QACzC,OAAO,IAAI,CAAC,UAAU,CAAC;IACzB,CAAC;IAED,MAAM,MAAM,GAAG,YAAY,CACzB,IAAI,CAAC,UAAU,EACf,IAAI,CAAC,eAAe,EACpB,IAAI,CAAC,kBAAkB,CACxB,CAAC;IACF,OAAO,iBAAiB,CAAC;QACvB,UAAU,EAAE,MAAM;QAClB,IAAI,EAAE,IAAI,CAAC,YAAY;QACvB,eAAe,EAAE,IAAI,CAAC,eAAe;QACrC,MAAM,EAAE,0BAA0B;KACnC,CAAC,CAAC;AACL,CAAC;AAED,MAAM,UAAU,yBAAyB,CAAC,IAKzC;IACC,MAAM,SAAS,GAAG,oBAAoB,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC;IACtD,MAAM,SAAS,GAAG,6BAA6B,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC;IAC/D,MAAM,aAAa,GAAG,IAAI,CAAC,OAAO,IAAI,CAAC,SAAS,IAAI,SAAS,CAAC,CAAC;IAE/D,IACE,IAAI,CAAC,OAAO;QACZ,CAAC,CACC,SAAS;YACT,SAAS;YACT,IAAI,CAAC,iBAAiB;YACtB,IAAI,CAAC,QAAQ,CAAC,WAAW,CAC1B,EACD,CAAC;QACD,MAAM,IAAI,8BAA8B,CAAC,IAAI,CAAC,eAAe,CAAC,CAAC;IACjE,CAAC;IAED,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,CAAC;QAClB,KAAK,CAAC,GAAG,IAAI,CAAC,eAAe,wBAAwB,CAAC,CAAC;IACzD,CAAC;SAAM,IAAI,aAAa,EAAE,CAAC;QACzB,KAAK,CAAC,GAAG,IAAI,CAAC,eAAe,iCAAiC,CAAC,CAAC;IAClE,CAAC;SAAM,IAAI,IAAI,CAAC,iBAAiB,EAAE,CAAC;QAClC,KAAK,CAAC,GAAG,IAAI,CAAC,eAAe,0CAA0C,CAAC,CAAC;IAC3E,CAAC;SAAM,CAAC;QACN,KAAK,CAAC,GAAG,IAAI,CAAC,eAAe,6CAA6C,CAAC,CAAC;IAC9E,CAAC;IAED,OAAO;QACL,aAAa;QACb,OAAO,EAAE,KAAK,EAAE,cAAc,EAAE,EAAE;YAChC,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,CAAC;gBAClB,OAAO,EAAE,CAAC;YACZ,CAAC;YAED,IAAI,UAAoC,CAAC;YACzC,IAAI,MAAc,CAAC;YACnB,IAAI,aAAa,EAAE,CAAC;gBAClB,UAAU,GAAG,cAAc,CAAC,kBAAkB,IAAI,EAAE,CAAC;gBACrD,MAAM,GAAG,IAAI,CAAC,eAAe,CAAC;YAChC,CAAC;iBAAM,IAAI,IAAI,CAAC,iBAAiB,EAAE,CAAC;gBAClC,UAAU,GAAG,MAAM,2BAA2B,CAAC;oBAC7C,WAAW,EAAE,cAAc,CAAC,WAAW;oBACvC,KAAK,EAAE,cAAc,CAAC,KAAK;oBAC3B,SAAS,EAAE,cAAc,CAAC,SAAS;oBACnC,QAAQ,EAAE,IAAI,CAAC,iBAAiB;oBAChC,IAAI,EAAE,cAAc,CAAC,eAAe;iBACrC,CAAC,CAAC;gBACH,MAAM,GAAG,mBAAmB,CAAC;YAC/B,CAAC;iBAAM,CAAC;gBACN,MAAM,QAAQ,GAAG,IAAI,CAAC,QAAQ,CAAC,WAAW,CAAC;gBAC3C,IAAI,CAAC,QAAQ,EAAE,CAAC;oBACd,MAAM,IAAI,8BAA8B,CAAC,IAAI,CAAC,eAAe,CAAC,CAAC;gBACjE,CAAC;gBACD,UAAU,GAAG,MAAM,sBAAsB,CAAC;oBACxC,QAAQ,EAAE,IAAI,CAAC,eAAe;oBAC9B,KAAK,EAAE,cAAc,CAAC,KAAK;oBAC3B,SAAS,EAAE,cAAc,CAAC,SAAS;oBACnC,IAAI,EAAE,cAAc,CAAC,eAAe;oBACpC,iBAAiB,EAAE,QAAQ;oBAC3B,WAAW,EAAE,cAAc,CAAC,WAAW;iBACxC,CAAC,CAAC;gBACH,MAAM,GAAG,GAAG,QAAQ,CAAC,QAAQ,CAAC,EAAE,IAAI,QAAQ,CAAC,OAAO,EAAE,CAAC;YACzD,CAAC;YAED,MAAM,SAAS,GAAG,iBAAiB,CAAC;gBAClC,UAAU;gBACV,IAAI,EAAE,cAAc,CAAC,eAAe;gBACpC,eAAe,EAAE,cAAc,CAAC,eAAe;gBAC/C,MAAM;aACP,CAAC,CAAC;YACH,OAAO;gBACL,UAAU,EAAE,iBAAiB,CAAC;oBAC5B,UAAU,EAAE,SAAS;oBACrB,eAAe,EAAE,cAAc,CAAC,eAAe;oBAC/C,YAAY,EAAE,cAAc,CAAC,YAAY;oBACzC,kBAAkB,EAAE,cAAc,CAAC,kBAAkB;oBACrD,eAAe,EAAE,cAAc,CAAC,eAAe;iBAChD,CAAC;aACH,CAAC;QACJ,CAAC;KACF,CAAC;AACJ,CAAC"}
|
|
@@ -0,0 +1,15 @@
|
|
|
1
|
+
import type { WordTimestamp } from "./timestamps.js";
|
|
2
|
+
export type TimestampFinalizationResult = {
|
|
3
|
+
readonly ok: true;
|
|
4
|
+
readonly timestamps: readonly WordTimestamp[];
|
|
5
|
+
} | {
|
|
6
|
+
readonly ok: false;
|
|
7
|
+
readonly reason: TimestampRejectionReason;
|
|
8
|
+
};
|
|
9
|
+
export type TimestampRejectionReason = "empty" | "invalid_timing" | "transcript_mismatch";
|
|
10
|
+
export declare function finalizeTimestamps(args: {
|
|
11
|
+
readonly audioDurationMs?: number;
|
|
12
|
+
readonly text: string;
|
|
13
|
+
readonly timestamps: readonly WordTimestamp[];
|
|
14
|
+
}): TimestampFinalizationResult;
|
|
15
|
+
//# sourceMappingURL=timestamp-finalization.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"timestamp-finalization.d.ts","sourceRoot":"","sources":["../src/timestamp-finalization.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,iBAAiB,CAAC;AAYrD,MAAM,MAAM,2BAA2B,GACnC;IACE,QAAQ,CAAC,EAAE,EAAE,IAAI,CAAC;IAClB,QAAQ,CAAC,UAAU,EAAE,SAAS,aAAa,EAAE,CAAC;CAC/C,GACD;IACE,QAAQ,CAAC,EAAE,EAAE,KAAK,CAAC;IACnB,QAAQ,CAAC,MAAM,EAAE,wBAAwB,CAAC;CAC3C,CAAC;AAEN,MAAM,MAAM,wBAAwB,GAChC,OAAO,GACP,gBAAgB,GAChB,qBAAqB,CAAC;AAqE1B,wBAAgB,kBAAkB,CAAC,IAAI,EAAE;IACvC,QAAQ,CAAC,eAAe,CAAC,EAAE,MAAM,CAAC;IAClC,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,UAAU,EAAE,SAAS,aAAa,EAAE,CAAC;CAC/C,GAAG,2BAA2B,CA6D9B"}
|
|
@@ -0,0 +1,111 @@
|
|
|
1
|
+
const CANONICAL_CHARACTERS = /[\p{L}\p{M}\p{N}]/gu;
|
|
2
|
+
const LEXICAL_CHARACTER = /[\p{L}\p{N}]/u;
|
|
3
|
+
const NON_WHITESPACE_RUN = /\S+/gu;
|
|
4
|
+
const TIMING_EPSILON_SECONDS = 1e-6;
|
|
5
|
+
function canonicalize(text) {
|
|
6
|
+
return (text.normalize("NFC").toLowerCase().match(CANONICAL_CHARACTERS) ?? []).join("");
|
|
7
|
+
}
|
|
8
|
+
function tokenizeSource(text) {
|
|
9
|
+
const ranges = [];
|
|
10
|
+
let leadingPunctuationStart;
|
|
11
|
+
for (const match of text.matchAll(NON_WHITESPACE_RUN)) {
|
|
12
|
+
const runText = match[0];
|
|
13
|
+
const start = match.index;
|
|
14
|
+
const end = start + runText.length;
|
|
15
|
+
if (LEXICAL_CHARACTER.test(runText)) {
|
|
16
|
+
ranges.push({ start: leadingPunctuationStart ?? start, end });
|
|
17
|
+
leadingPunctuationStart = undefined;
|
|
18
|
+
continue;
|
|
19
|
+
}
|
|
20
|
+
const previous = ranges.at(-1);
|
|
21
|
+
if (previous) {
|
|
22
|
+
previous.end = end;
|
|
23
|
+
}
|
|
24
|
+
else {
|
|
25
|
+
leadingPunctuationStart ??= start;
|
|
26
|
+
}
|
|
27
|
+
}
|
|
28
|
+
return ranges.map(({ start, end }) => {
|
|
29
|
+
const tokenText = text.slice(start, end);
|
|
30
|
+
return { canonical: canonicalize(tokenText), text: tokenText };
|
|
31
|
+
});
|
|
32
|
+
}
|
|
33
|
+
function hasInvalidTiming(timestamps, audioDurationMs) {
|
|
34
|
+
const durationSeconds = audioDurationMs != null && audioDurationMs > 0
|
|
35
|
+
? audioDurationMs / 1000
|
|
36
|
+
: undefined;
|
|
37
|
+
const durationTolerance = durationSeconds == null
|
|
38
|
+
? undefined
|
|
39
|
+
: Math.max(0.25, durationSeconds * 0.05);
|
|
40
|
+
let previousEnd = 0;
|
|
41
|
+
for (const [index, timestamp] of timestamps.entries()) {
|
|
42
|
+
if (!(Number.isFinite(timestamp.start) && Number.isFinite(timestamp.end)) ||
|
|
43
|
+
timestamp.start < 0 ||
|
|
44
|
+
timestamp.end < 0 ||
|
|
45
|
+
timestamp.end + TIMING_EPSILON_SECONDS < timestamp.start ||
|
|
46
|
+
(index > 0 && timestamp.start + TIMING_EPSILON_SECONDS < previousEnd) ||
|
|
47
|
+
(durationSeconds != null &&
|
|
48
|
+
durationTolerance != null &&
|
|
49
|
+
timestamp.end > durationSeconds + durationTolerance)) {
|
|
50
|
+
return true;
|
|
51
|
+
}
|
|
52
|
+
previousEnd = timestamp.end;
|
|
53
|
+
}
|
|
54
|
+
return false;
|
|
55
|
+
}
|
|
56
|
+
export function finalizeTimestamps(args) {
|
|
57
|
+
const sourceTokens = tokenizeSource(args.text);
|
|
58
|
+
if (sourceTokens.length === 0) {
|
|
59
|
+
return args.timestamps.length === 0
|
|
60
|
+
? { ok: true, timestamps: [] }
|
|
61
|
+
: { ok: false, reason: "transcript_mismatch" };
|
|
62
|
+
}
|
|
63
|
+
if (args.timestamps.length === 0) {
|
|
64
|
+
return { ok: false, reason: "empty" };
|
|
65
|
+
}
|
|
66
|
+
if (args.timestamps.some(({ text }) => !LEXICAL_CHARACTER.test(text)) ||
|
|
67
|
+
args.timestamps.some(({ text }) => canonicalize(text).length === 0)) {
|
|
68
|
+
return { ok: false, reason: "transcript_mismatch" };
|
|
69
|
+
}
|
|
70
|
+
if (hasInvalidTiming(args.timestamps, args.audioDurationMs)) {
|
|
71
|
+
return { ok: false, reason: "invalid_timing" };
|
|
72
|
+
}
|
|
73
|
+
const sourceCanonical = sourceTokens
|
|
74
|
+
.map(({ canonical }) => canonical)
|
|
75
|
+
.join("");
|
|
76
|
+
const providerCanonical = args.timestamps
|
|
77
|
+
.map(({ text }) => canonicalize(text))
|
|
78
|
+
.join("");
|
|
79
|
+
if (sourceCanonical !== providerCanonical) {
|
|
80
|
+
return { ok: false, reason: "transcript_mismatch" };
|
|
81
|
+
}
|
|
82
|
+
const providerBoundaries = new Map();
|
|
83
|
+
let providerOffset = 0;
|
|
84
|
+
for (const [index, timestamp] of args.timestamps.entries()) {
|
|
85
|
+
providerOffset += [...canonicalize(timestamp.text)].length;
|
|
86
|
+
providerBoundaries.set(providerOffset, index);
|
|
87
|
+
}
|
|
88
|
+
const projected = [];
|
|
89
|
+
let sourceOffset = 0;
|
|
90
|
+
let firstProviderIndex = 0;
|
|
91
|
+
for (const sourceToken of sourceTokens) {
|
|
92
|
+
sourceOffset += [...sourceToken.canonical].length;
|
|
93
|
+
const lastProviderIndex = providerBoundaries.get(sourceOffset);
|
|
94
|
+
if (lastProviderIndex == null || lastProviderIndex < firstProviderIndex) {
|
|
95
|
+
return { ok: false, reason: "transcript_mismatch" };
|
|
96
|
+
}
|
|
97
|
+
const first = args.timestamps[firstProviderIndex];
|
|
98
|
+
const last = args.timestamps[lastProviderIndex];
|
|
99
|
+
if (!(first && last)) {
|
|
100
|
+
return { ok: false, reason: "transcript_mismatch" };
|
|
101
|
+
}
|
|
102
|
+
projected.push({
|
|
103
|
+
text: sourceToken.text,
|
|
104
|
+
start: first.start,
|
|
105
|
+
end: last.end,
|
|
106
|
+
});
|
|
107
|
+
firstProviderIndex = lastProviderIndex + 1;
|
|
108
|
+
}
|
|
109
|
+
return { ok: true, timestamps: projected };
|
|
110
|
+
}
|
|
111
|
+
//# sourceMappingURL=timestamp-finalization.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"timestamp-finalization.js","sourceRoot":"","sources":["../src/timestamp-finalization.ts"],"names":[],"mappings":"AAEA,MAAM,oBAAoB,GAAG,qBAAqB,CAAC;AACnD,MAAM,iBAAiB,GAAG,eAAe,CAAC;AAC1C,MAAM,kBAAkB,GAAG,OAAO,CAAC;AACnC,MAAM,sBAAsB,GAAG,IAAI,CAAC;AAsBpC,SAAS,YAAY,CAAC,IAAY;IAChC,OAAO,CACL,IAAI,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,WAAW,EAAE,CAAC,KAAK,CAAC,oBAAoB,CAAC,IAAI,EAAE,CACtE,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;AACb,CAAC;AAED,SAAS,cAAc,CAAC,IAAY;IAClC,MAAM,MAAM,GAAqC,EAAE,CAAC;IACpD,IAAI,uBAA2C,CAAC;IAEhD,KAAK,MAAM,KAAK,IAAI,IAAI,CAAC,QAAQ,CAAC,kBAAkB,CAAC,EAAE,CAAC;QACtD,MAAM,OAAO,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC;QACzB,MAAM,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC;QAC1B,MAAM,GAAG,GAAG,KAAK,GAAG,OAAO,CAAC,MAAM,CAAC;QAEnC,IAAI,iBAAiB,CAAC,IAAI,CAAC,OAAO,CAAC,EAAE,CAAC;YACpC,MAAM,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,uBAAuB,IAAI,KAAK,EAAE,GAAG,EAAE,CAAC,CAAC;YAC9D,uBAAuB,GAAG,SAAS,CAAC;YACpC,SAAS;QACX,CAAC;QAED,MAAM,QAAQ,GAAG,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;QAC/B,IAAI,QAAQ,EAAE,CAAC;YACb,QAAQ,CAAC,GAAG,GAAG,GAAG,CAAC;QACrB,CAAC;aAAM,CAAC;YACN,uBAAuB,KAAK,KAAK,CAAC;QACpC,CAAC;IACH,CAAC;IAED,OAAO,MAAM,CAAC,GAAG,CAAC,CAAC,EAAE,KAAK,EAAE,GAAG,EAAE,EAAE,EAAE;QACnC,MAAM,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,KAAK,EAAE,GAAG,CAAC,CAAC;QACzC,OAAO,EAAE,SAAS,EAAE,YAAY,CAAC,SAAS,CAAC,EAAE,IAAI,EAAE,SAAS,EAAE,CAAC;IACjE,CAAC,CAAC,CAAC;AACL,CAAC;AAED,SAAS,gBAAgB,CACvB,UAAoC,EACpC,eAAmC;IAEnC,MAAM,eAAe,GACnB,eAAe,IAAI,IAAI,IAAI,eAAe,GAAG,CAAC;QAC5C,CAAC,CAAC,eAAe,GAAG,IAAI;QACxB,CAAC,CAAC,SAAS,CAAC;IAChB,MAAM,iBAAiB,GACrB,eAAe,IAAI,IAAI;QACrB,CAAC,CAAC,SAAS;QACX,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,eAAe,GAAG,IAAI,CAAC,CAAC;IAE7C,IAAI,WAAW,GAAG,CAAC,CAAC;IACpB,KAAK,MAAM,CAAC,KAAK,EAAE,SAAS,CAAC,IAAI,UAAU,CAAC,OAAO,EAAE,EAAE,CAAC;QACtD,IACE,CAAC,CAAC,MAAM,CAAC,QAAQ,CAAC,SAAS,CAAC,KAAK,CAAC,IAAI,MAAM,CAAC,QAAQ,CAAC,SAAS,CAAC,GAAG,CAAC,CAAC;YACrE,SAAS,CAAC,KAAK,GAAG,CAAC;YACnB,SAAS,CAAC,GAAG,GAAG,CAAC;YACjB,SAAS,CAAC,GAAG,GAAG,sBAAsB,GAAG,SAAS,CAAC,KAAK;YACxD,CAAC,KAAK,GAAG,CAAC,IAAI,SAAS,CAAC,KAAK,GAAG,sBAAsB,GAAG,WAAW,CAAC;YACrE,CAAC,eAAe,IAAI,IAAI;gBACtB,iBAAiB,IAAI,IAAI;gBACzB,SAAS,CAAC,GAAG,GAAG,eAAe,GAAG,iBAAiB,CAAC,EACtD,CAAC;YACD,OAAO,IAAI,CAAC;QACd,CAAC;QACD,WAAW,GAAG,SAAS,CAAC,GAAG,CAAC;IAC9B,CAAC;IACD,OAAO,KAAK,CAAC;AACf,CAAC;AAED,MAAM,UAAU,kBAAkB,CAAC,IAIlC;IACC,MAAM,YAAY,GAAG,cAAc,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IAE/C,IAAI,YAAY,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QAC9B,OAAO,IAAI,CAAC,UAAU,CAAC,MAAM,KAAK,CAAC;YACjC,CAAC,CAAC,EAAE,EAAE,EAAE,IAAI,EAAE,UAAU,EAAE,EAAE,EAAE;YAC9B,CAAC,CAAC,EAAE,EAAE,EAAE,KAAK,EAAE,MAAM,EAAE,qBAAqB,EAAE,CAAC;IACnD,CAAC;IACD,IAAI,IAAI,CAAC,UAAU,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QACjC,OAAO,EAAE,EAAE,EAAE,KAAK,EAAE,MAAM,EAAE,OAAO,EAAE,CAAC;IACxC,CAAC;IACD,IACE,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC,CAAC,EAAE,IAAI,EAAE,EAAE,EAAE,CAAC,CAAC,iBAAiB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QACjE,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC,CAAC,EAAE,IAAI,EAAE,EAAE,EAAE,CAAC,YAAY,CAAC,IAAI,CAAC,CAAC,MAAM,KAAK,CAAC,CAAC,EACnE,CAAC;QACD,OAAO,EAAE,EAAE,EAAE,KAAK,EAAE,MAAM,EAAE,qBAAqB,EAAE,CAAC;IACtD,CAAC;IACD,IAAI,gBAAgB,CAAC,IAAI,CAAC,UAAU,EAAE,IAAI,CAAC,eAAe,CAAC,EAAE,CAAC;QAC5D,OAAO,EAAE,EAAE,EAAE,KAAK,EAAE,MAAM,EAAE,gBAAgB,EAAE,CAAC;IACjD,CAAC;IAED,MAAM,eAAe,GAAG,YAAY;SACjC,GAAG,CAAC,CAAC,EAAE,SAAS,EAAE,EAAE,EAAE,CAAC,SAAS,CAAC;SACjC,IAAI,CAAC,EAAE,CAAC,CAAC;IACZ,MAAM,iBAAiB,GAAG,IAAI,CAAC,UAAU;SACtC,GAAG,CAAC,CAAC,EAAE,IAAI,EAAE,EAAE,EAAE,CAAC,YAAY,CAAC,IAAI,CAAC,CAAC;SACrC,IAAI,CAAC,EAAE,CAAC,CAAC;IACZ,IAAI,eAAe,KAAK,iBAAiB,EAAE,CAAC;QAC1C,OAAO,EAAE,EAAE,EAAE,KAAK,EAAE,MAAM,EAAE,qBAAqB,EAAE,CAAC;IACtD,CAAC;IAED,MAAM,kBAAkB,GAAG,IAAI,GAAG,EAAkB,CAAC;IACrD,IAAI,cAAc,GAAG,CAAC,CAAC;IACvB,KAAK,MAAM,CAAC,KAAK,EAAE,SAAS,CAAC,IAAI,IAAI,CAAC,UAAU,CAAC,OAAO,EAAE,EAAE,CAAC;QAC3D,cAAc,IAAI,CAAC,GAAG,YAAY,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC,MAAM,CAAC;QAC3D,kBAAkB,CAAC,GAAG,CAAC,cAAc,EAAE,KAAK,CAAC,CAAC;IAChD,CAAC;IAED,MAAM,SAAS,GAAoB,EAAE,CAAC;IACtC,IAAI,YAAY,GAAG,CAAC,CAAC;IACrB,IAAI,kBAAkB,GAAG,CAAC,CAAC;IAC3B,KAAK,MAAM,WAAW,IAAI,YAAY,EAAE,CAAC;QACvC,YAAY,IAAI,CAAC,GAAG,WAAW,CAAC,SAAS,CAAC,CAAC,MAAM,CAAC;QAClD,MAAM,iBAAiB,GAAG,kBAAkB,CAAC,GAAG,CAAC,YAAY,CAAC,CAAC;QAC/D,IAAI,iBAAiB,IAAI,IAAI,IAAI,iBAAiB,GAAG,kBAAkB,EAAE,CAAC;YACxE,OAAO,EAAE,EAAE,EAAE,KAAK,EAAE,MAAM,EAAE,qBAAqB,EAAE,CAAC;QACtD,CAAC;QACD,MAAM,KAAK,GAAG,IAAI,CAAC,UAAU,CAAC,kBAAkB,CAAC,CAAC;QAClD,MAAM,IAAI,GAAG,IAAI,CAAC,UAAU,CAAC,iBAAiB,CAAC,CAAC;QAChD,IAAI,CAAC,CAAC,KAAK,IAAI,IAAI,CAAC,EAAE,CAAC;YACrB,OAAO,EAAE,EAAE,EAAE,KAAK,EAAE,MAAM,EAAE,qBAAqB,EAAE,CAAC;QACtD,CAAC;QACD,SAAS,CAAC,IAAI,CAAC;YACb,IAAI,EAAE,WAAW,CAAC,IAAI;YACtB,KAAK,EAAE,KAAK,CAAC,KAAK;YAClB,GAAG,EAAE,IAAI,CAAC,GAAG;SACd,CAAC,CAAC;QACH,kBAAkB,GAAG,iBAAiB,GAAG,CAAC,CAAC;IAC7C,CAAC;IAED,OAAO,EAAE,EAAE,EAAE,IAAI,EAAE,UAAU,EAAE,SAAS,EAAE,CAAC;AAC7C,CAAC"}
|
|
@@ -0,0 +1,10 @@
|
|
|
1
|
+
import type { WordTimestamp } from "./timestamps.js";
|
|
2
|
+
export interface TimestampProvider {
|
|
3
|
+
align(input: {
|
|
4
|
+
readonly abortSignal?: AbortSignal;
|
|
5
|
+
readonly audio: Uint8Array;
|
|
6
|
+
readonly mediaType: string;
|
|
7
|
+
readonly text: string;
|
|
8
|
+
}): Promise<readonly WordTimestamp[]>;
|
|
9
|
+
}
|
|
10
|
+
//# sourceMappingURL=timestamp-provider.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"timestamp-provider.d.ts","sourceRoot":"","sources":["../src/timestamp-provider.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,iBAAiB,CAAC;AAErD,MAAM,WAAW,iBAAiB;IAChC,KAAK,CAAC,KAAK,EAAE;QACX,QAAQ,CAAC,WAAW,CAAC,EAAE,WAAW,CAAC;QACnC,QAAQ,CAAC,KAAK,EAAE,UAAU,CAAC;QAC3B,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;QAC3B,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;KACvB,GAAG,OAAO,CAAC,SAAS,aAAa,EAAE,CAAC,CAAC;CACvC"}
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"timestamp-provider.js","sourceRoot":"","sources":["../src/timestamp-provider.ts"],"names":[],"mappings":""}
|
package/dist/types.d.ts
CHANGED
|
@@ -1,6 +1,7 @@
|
|
|
1
1
|
import type { AudioOutput } from "./audio-output.js";
|
|
2
2
|
import type { PronunciationsInput } from "./pronunciations/types.js";
|
|
3
3
|
import type { ResolvedModel, Voice } from "./speech-provider.js";
|
|
4
|
+
import type { TimestampProvider } from "./timestamp-provider.js";
|
|
4
5
|
export type { AudioOutput, AudioOutputFormat } from "./audio-output.js";
|
|
5
6
|
export type { CaptionFormat, CaptionsOptions } from "./captions.js";
|
|
6
7
|
export type { ClonedVoice, CloneVoiceOptions, SpeechProviderFactory, VoiceSample, } from "./clone-voice.js";
|
|
@@ -24,9 +25,10 @@ export type { OpenAISpeechProviderConfig } from "./providers/openai/index.js";
|
|
|
24
25
|
export type { ResembleSpeechProviderConfig } from "./providers/resemble/index.js";
|
|
25
26
|
export type { XaiSpeechProviderConfig } from "./providers/xai/index.js";
|
|
26
27
|
export type { Feature, ModelInfo, ResolvedModel, SpeechProvider, Voice, } from "./speech-provider.js";
|
|
27
|
-
export type { ConversationResult, GeneratedAudioFile, SpeechResult, } from "./speech-result.js";
|
|
28
|
+
export type { ConversationResult, ConversationResultWithTimestamps, GeneratedAudioFile, SpeechResult, SpeechResultWithTimestamps, } from "./speech-result.js";
|
|
28
29
|
export type { ResolvedSTTModel, SpeechToTextProvider, STTModelInfo, } from "./speech-to-text-provider.js";
|
|
29
30
|
export type { StreamSpeechResult } from "./stream-speech-result.js";
|
|
31
|
+
export type { TimestampProvider } from "./timestamp-provider.js";
|
|
30
32
|
export type { ConversationWordTimestamp, WordTimestamp, } from "./timestamps.js";
|
|
31
33
|
export type { TurnTimestamp } from "./turns.js";
|
|
32
34
|
export interface GenerateSpeechOptions<V extends Voice = Voice, M extends string | ResolvedModel<V> = string | ResolvedModel<V>> {
|
|
@@ -42,6 +44,7 @@ export interface GenerateSpeechOptions<V extends Voice = Voice, M extends string
|
|
|
42
44
|
providerOptions?: Record<string, unknown>;
|
|
43
45
|
speed?: number;
|
|
44
46
|
text: string;
|
|
47
|
+
timestampProvider?: TimestampProvider;
|
|
45
48
|
timestamps?: boolean;
|
|
46
49
|
voice: V;
|
|
47
50
|
volumeDbfs?: number;
|
package/dist/types.d.ts.map
CHANGED
|
@@ -1 +1 @@
|
|
|
1
|
-
{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AACrD,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,2BAA2B,CAAC;AACrE,OAAO,KAAK,EAAE,aAAa,EAAE,KAAK,EAAE,MAAM,sBAAsB,CAAC;AAEjE,YAAY,EAAE,WAAW,EAAE,iBAAiB,EAAE,MAAM,mBAAmB,CAAC;AACxE,YAAY,EAAE,aAAa,EAAE,eAAe,EAAE,MAAM,eAAe,CAAC;AACpE,YAAY,EACV,WAAW,EACX,iBAAiB,EACjB,qBAAqB,EACrB,WAAW,GACZ,MAAM,kBAAkB,CAAC;AAC1B,YAAY,EACV,gBAAgB,EAChB,2BAA2B,GAC5B,MAAM,yBAAyB,CAAC;AACjC,YAAY,EACV,aAAa,EACb,kBAAkB,EAClB,kBAAkB,GACnB,MAAM,mBAAmB,CAAC;AAC3B,YAAY,EAAE,cAAc,EAAE,MAAM,eAAe,CAAC;AACpD,YAAY,EACV,aAAa,EACb,mBAAmB,GACpB,MAAM,2BAA2B,CAAC;AACnC,YAAY,EAAE,4BAA4B,EAAE,MAAM,+BAA+B,CAAC;AAClF,YAAY,EAAE,4BAA4B,EAAE,MAAM,+BAA+B,CAAC;AAClF,YAAY,EAAE,8BAA8B,EAAE,MAAM,iCAAiC,CAAC;AACtF,YAAY,EAAE,uBAAuB,EAAE,MAAM,0BAA0B,CAAC;AACxE,YAAY,EAAE,6BAA6B,EAAE,MAAM,iCAAiC,CAAC;AACrF,YAAY,EAAE,2BAA2B,EAAE,MAAM,8BAA8B,CAAC;AAChF,YAAY,EAAE,0BAA0B,EAAE,MAAM,6BAA6B,CAAC;AAC9E,YAAY,EAAE,2BAA2B,EAAE,MAAM,8BAA8B,CAAC;AAChF,YAAY,EAAE,wBAAwB,EAAE,MAAM,2BAA2B,CAAC;AAC1E,YAAY,EAAE,2BAA2B,EAAE,MAAM,8BAA8B,CAAC;AAChF,YAAY,EAAE,2BAA2B,EAAE,MAAM,8BAA8B,CAAC;AAChF,YAAY,EAAE,wBAAwB,EAAE,MAAM,2BAA2B,CAAC;AAC1E,YAAY,EAAE,0BAA0B,EAAE,MAAM,6BAA6B,CAAC;AAC9E,YAAY,EAAE,4BAA4B,EAAE,MAAM,+BAA+B,CAAC;AAClF,YAAY,EAAE,uBAAuB,EAAE,MAAM,0BAA0B,CAAC;AACxE,YAAY,EACV,OAAO,EACP,SAAS,EACT,aAAa,EACb,cAAc,EACd,KAAK,GACN,MAAM,sBAAsB,CAAC;AAC9B,YAAY,EACV,kBAAkB,EAClB,kBAAkB,EAClB,YAAY,
|
|
1
|
+
{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AACrD,OAAO,KAAK,EAAE,mBAAmB,EAAE,MAAM,2BAA2B,CAAC;AACrE,OAAO,KAAK,EAAE,aAAa,EAAE,KAAK,EAAE,MAAM,sBAAsB,CAAC;AACjE,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,yBAAyB,CAAC;AAEjE,YAAY,EAAE,WAAW,EAAE,iBAAiB,EAAE,MAAM,mBAAmB,CAAC;AACxE,YAAY,EAAE,aAAa,EAAE,eAAe,EAAE,MAAM,eAAe,CAAC;AACpE,YAAY,EACV,WAAW,EACX,iBAAiB,EACjB,qBAAqB,EACrB,WAAW,GACZ,MAAM,kBAAkB,CAAC;AAC1B,YAAY,EACV,gBAAgB,EAChB,2BAA2B,GAC5B,MAAM,yBAAyB,CAAC;AACjC,YAAY,EACV,aAAa,EACb,kBAAkB,EAClB,kBAAkB,GACnB,MAAM,mBAAmB,CAAC;AAC3B,YAAY,EAAE,cAAc,EAAE,MAAM,eAAe,CAAC;AACpD,YAAY,EACV,aAAa,EACb,mBAAmB,GACpB,MAAM,2BAA2B,CAAC;AACnC,YAAY,EAAE,4BAA4B,EAAE,MAAM,+BAA+B,CAAC;AAClF,YAAY,EAAE,4BAA4B,EAAE,MAAM,+BAA+B,CAAC;AAClF,YAAY,EAAE,8BAA8B,EAAE,MAAM,iCAAiC,CAAC;AACtF,YAAY,EAAE,uBAAuB,EAAE,MAAM,0BAA0B,CAAC;AACxE,YAAY,EAAE,6BAA6B,EAAE,MAAM,iCAAiC,CAAC;AACrF,YAAY,EAAE,2BAA2B,EAAE,MAAM,8BAA8B,CAAC;AAChF,YAAY,EAAE,0BAA0B,EAAE,MAAM,6BAA6B,CAAC;AAC9E,YAAY,EAAE,2BAA2B,EAAE,MAAM,8BAA8B,CAAC;AAChF,YAAY,EAAE,wBAAwB,EAAE,MAAM,2BAA2B,CAAC;AAC1E,YAAY,EAAE,2BAA2B,EAAE,MAAM,8BAA8B,CAAC;AAChF,YAAY,EAAE,2BAA2B,EAAE,MAAM,8BAA8B,CAAC;AAChF,YAAY,EAAE,wBAAwB,EAAE,MAAM,2BAA2B,CAAC;AAC1E,YAAY,EAAE,0BAA0B,EAAE,MAAM,6BAA6B,CAAC;AAC9E,YAAY,EAAE,4BAA4B,EAAE,MAAM,+BAA+B,CAAC;AAClF,YAAY,EAAE,uBAAuB,EAAE,MAAM,0BAA0B,CAAC;AACxE,YAAY,EACV,OAAO,EACP,SAAS,EACT,aAAa,EACb,cAAc,EACd,KAAK,GACN,MAAM,sBAAsB,CAAC;AAC9B,YAAY,EACV,kBAAkB,EAClB,gCAAgC,EAChC,kBAAkB,EAClB,YAAY,EACZ,0BAA0B,GAC3B,MAAM,oBAAoB,CAAC;AAC5B,YAAY,EACV,gBAAgB,EAChB,oBAAoB,EACpB,YAAY,GACb,MAAM,8BAA8B,CAAC;AACtC,YAAY,EAAE,kBAAkB,EAAE,MAAM,2BAA2B,CAAC;AACpE,YAAY,EAAE,iBAAiB,EAAE,MAAM,yBAAyB,CAAC;AACjE,YAAY,EACV,yBAAyB,EACzB,aAAa,GACd,MAAM,iBAAiB,CAAC;AACzB,YAAY,EAAE,aAAa,EAAE,MAAM,YAAY,CAAC;AAEhD,MAAM,WAAW,qBAAqB,CACpC,CAAC,SAAS,KAAK,GAAG,KAAK,EACvB,CAAC,SAAS,MAAM,GAAG,aAAa,CAAC,CAAC,CAAC,GAAG,MAAM,GAAG,aAAa,CAAC,CAAC,CAAC;IAE/D,WAAW,CAAC,EAAE,WAAW,CAAC;IAC1B,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC;IAEjC,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,KAAK,EAAE,CAAC,CAAC;IACT,MAAM,CAAC,EAAE,WAAW,CAAC;IACrB,cAAc,CAAC,EAAE,mBAAmB,CAAC;IACrC,eAAe,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAE1C,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,IAAI,EAAE,MAAM,CAAC;IACb,iBAAiB,CAAC,EAAE,iBAAiB,CAAC;IACtC,UAAU,CAAC,EAAE,OAAO,CAAC;IACrB,KAAK,EAAE,CAAC,CAAC;IACT,UAAU,CAAC,EAAE,MAAM,CAAC;CACrB"}
|
package/package.json
CHANGED
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"default-stt-fallback.d.ts","sourceRoot":"","sources":["../src/default-stt-fallback.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,8BAA8B,CAAC;AAKrE,wBAAsB,qBAAqB,IAAI,OAAO,CAAC,gBAAgB,CAAC,CAOvE"}
|
|
@@ -1,11 +0,0 @@
|
|
|
1
|
-
let cached;
|
|
2
|
-
// Dynamic import keeps the OpenAI provider out of bundles for callers who never trigger the STT fallback.
|
|
3
|
-
export async function getDefaultSTTFallback() {
|
|
4
|
-
if (cached) {
|
|
5
|
-
return cached;
|
|
6
|
-
}
|
|
7
|
-
const { createOpenAI } = await import("./providers/openai/index.js");
|
|
8
|
-
cached = createOpenAI().stt("whisper-1");
|
|
9
|
-
return cached;
|
|
10
|
-
}
|
|
11
|
-
//# sourceMappingURL=default-stt-fallback.js.map
|
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"default-stt-fallback.js","sourceRoot":"","sources":["../src/default-stt-fallback.ts"],"names":[],"mappings":"AAEA,IAAI,MAAoC,CAAC;AAEzC,0GAA0G;AAC1G,MAAM,CAAC,KAAK,UAAU,qBAAqB;IACzC,IAAI,MAAM,EAAE,CAAC;QACX,OAAO,MAAM,CAAC;IAChB,CAAC;IACD,MAAM,EAAE,YAAY,EAAE,GAAG,MAAM,MAAM,CAAC,6BAA6B,CAAC,CAAC;IACrE,MAAM,GAAG,YAAY,EAAE,CAAC,GAAG,CAAC,WAAW,CAAC,CAAC;IACzC,OAAO,MAAM,CAAC;AAChB,CAAC"}
|